How to decay your learning rate

Aitor Lewkowycz

Keywords: learning rates, hyperparameter tuning, schedules

Abstract: Complex learning rate schedules have become an integral part of deep learning. We find empirically that common fine-tuned schedules decay the learning rate after the weight norm bounces. This leads to the proposal of ABEL: an automatic scheduler which decays the learning rate by keeping track of the weight norm. ABEL's performance matches that of tuned schedules, is more robust with respect to its parameters and does not depend on the time budget. Through extensive experiments in vision, NLP, and RL, we show that if the weight norm does not bounce, we can simplify schedules even further with no loss in performance. In such cases, a complex schedule has similar performance to a constant learning rate with a decay at the end of training.

One-sentence Summary: We study when it is beneficial to decay the learning rate and how to do it.

Community Implementations: [ 3 code implementations](https://www.catalyzex.com/paper/arxiv:2103.12682/code)